Back to skill

Security audit

kitbay

Security checks for vulnerabilities and agentic risk

Overview

The verdict is benign with medium confidence: scanners found no executable files or AIG issues, but flagged references to client configuration locations and sending task descriptions to Kitbay during search.

Before installing, consider whether searches could include private project details: the scanner reports they may be sent to Kitbay. Keep search terms generic and omit confidential information. The scan covered all reported components, but the benign verdict does not mean risk-free.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Agent Config Directory Access

High
Category
Agent Snooping
Confidence
90% confidence
Finding

Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.

Content

Scanner excerpt · reference/clients.md (reported line 70)May include surrounding context.

md
## Codex

- MCP config: ~/.codex/config.toml (default); .codex/config.toml in trusted projects ([source](https://learn.chatgpt.com/docs/extend/mcp?surface=cli))
- Skills: Repository: $CWD/.agents/skills, $CWD/../.agents/skills, $REPO_ROOT/.agents/skills; user: $HOME/.agents/skills; admin: /etc/codex/skills; system: bundled with Codex ([source](https://learn.chatgpt.com/docs/build-skills))
- Checked: 2026-11-14

Agent Config Directory Access

High
Category
Agent Snooping
Confidence
90% confidence
Finding

Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.

Content

Scanner excerpt · reference/clients.md (reported line 74)May include surrounding context.

  • Skills: Repository: $CWD/.agents/skills, $CWD/../.agents/skills, $REPO_ROOT/.agents/skills; user: $HOME/.agents/skills; admin: /etc/codex/skills; system: bundled with Codex (source)
  • Checked: 2026-11-14

Connect Kitbay (in ~/.codex/config.toml):

toml
[mcp_servers.kitbay]

Agent Config Directory Access

High
Category
Agent Snooping
Confidence
90% confidence
Finding

Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.

Content

Scanner excerpt · reference/clients.md (reported line 96)May include surrounding context.

md
## Gemini CLI

- MCP config: User: ~/.gemini/settings.json; project: .gemini/settings.json ([source](https://geminicli.com/docs/tools/mcp-server/))
- Skills: User: ~/.gemini/skills/ or ~/.agents/skills/; workspace: .gemini/skills/ or .agents/skills/; also bundled in extensions and built-in ([source](https://geminicli.com/docs/cli/skills/))
- Checked: 2026-11-14

The search workflow sends user-derived tool names and task descriptions to Kitbay without a privacy warning.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The workflow directs the agent to send user-derived task queries to an external directory without requiring sensitive-context filtering or disclosure before transmission. Ordinary public tool names present little risk, but task descriptions can contain confidential project names or internal requirements, especially when the skill activates because the agent needs a missing tool. The installation permission requirement does not cover these earlier search disclosures.

Content

No source excerpt is available for this finding.